autoconf-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Autotest: make test dir writable before removing


From: Stepan Kasal
Subject: Re: Autotest: make test dir writable before removing
Date: Tue, 4 Oct 2005 11:15:01 +0200
User-agent: Mutt/1.4.1i

Hello,

On Mon, Oct 03, 2005 at 12:41:04PM -0700, Paul Eggert wrote:
> I was worried about the case where it wasn't erased, and where the
> test contained a symlink to /usr/bin or something like that.
> Admittedly I'm waving my hands a bit here; I don't know exactly what
> I'm worried about.

I cannot imagine a real danger with links.
"mount --bind" would be dangerous, but I don't this Autotest test suties
should excersise it.  (It would be dangerous even without chmod -R.)

I'm afraid I rushed in too quickly when I decided to commit the patch.

> > I guess we have to put in the trap in order to provide backward
> > compatibility with 2.59, but only for that reason.
> 
> Yes, that's right.

First, I have to admit that I don't understand this, so I cannot fix it.

> Here are some other problems with that patch.
> 
> First, there are two other instances of the problem in general.m4:
> 
>       rm -f -r "$at_suite_dir" "$at_suite_log"
> 
>         $at_debug_p || rm -f -r $at_group_dir
> 
> Second, chmod u+w doesn't suffice for recursive removals.  You would
> need at least chmod u+rwx.  The 'r' is needed so that rm can read the
> directory to find out what its subfiles are.  The 'x' is needed so
> that rm can search the directory to remove its subfiles.

But I understand this, so I implemented it.  I committed another patch,
as attached to this mail.

Have a nice day,
        Stepan Kasal

Attachment: autoconf-20051004-at-init.patch
Description: Text document


reply via email to

[Prev in Thread] Current Thread [Next in Thread]